Gigabyte EasyTune6 shows different values for memory slots.

dmavro

Distinguished
Feb 11, 2010
223
5
18,685
I was looking at EasyTune6 on my system and noticed discrepancies between all of the memory slots. Is it normal for slot#1 ,#2,and #3 to have different frequency's, Ras# to Cas#'s , Ras# precharge, and Tras #'s all under the Timings Table. I also noticed that slot#1 and #3 show max bandwidth as PC3-10700H(667 mhz) and slot #2 shows PC3-17300H(1083 MHZ) and the part number's for each slot are also different. I purchased a 6 gb kit from corsair and assumed all part numbers should be the same. I have not overclocked manually, but i did set performance enhance to Extreme in the Bios. I just want to know if this is normal or if i have some issue somewhere in my bios or even a problem with my actual memory. Thanx in advance.
 
Issue resolved....i shut off extreme in bios and everything is the same for each slot now.... very odd?..... Does anyone have any recommended bios settings for the Ga-x58a-ud5 mobo.
 


Issue with RAM chips, is that some do better overclocking than others. Even from the same bin. What probably happened is that one chip took the high motherboard settings more gracefully than the other.

If you want your system life to be good, leave it as it is running.
 
i shut off extreme in bios and everything is the same for each slot now
If, by this, you mean you disabled XMP - I'd consider this an 'RMAable offense', and deal with it now, before going any further... (And, trust me, I don't take RMAs lightly - several times here, I have ranted for a page or two about 'useless' RMAs costing everyone money!) It certainly looks, from the symptoms, like you've somehow gotten a 'mis-matched' set of DIMMs.

You should be able to confirm, by D/Ling and running SPDTool 0.63; if you click on the 'Read' menu item, it should give you a list of your DIMMs - the part number for one of mine is circled:
0274d.jpg

...be very careful with this program - it can also write SPDs, which is typically not a good thing!

Once confirmed, contact Corsair here...

I can give you a procedure for 'sneaking around' this, but you are much, much better off getting it dealt with properly before trying to 'crank 'er up'!!

 


I already had X.M.P disabled. What i actually changed this time was performance enhance setting..... from Extreme to Standard.
 
I finally got a chance to use SPDTool 0.63 and all part numbers are the same. But i noticed today that easy tune is showing different values for max bandwidth again and some of the other values are different also. I havent changed anything since i disabled bios feature last week. Do u think its just easytune issue or do i possibly have bad ram. Computer has been acting odd for a few weeks now. I use pc to trade and once or twice a day I lose the ability to do anything on it. I just get a blue circle spinning for mouse cursor. Everything keeps updating. I just cant do anything for around a minute.
 
I wouldn't doubt that EasyTune is part of the problem, and will never be 'part of the solution' - you can see my opinion about running it on a 24/7 basis toward the end of Memory Part IV - "Tweaking and tuning", under the MemSet/CPU-Tweaker, EasyTune(X??), OverDrive, or BIOS: which should I use? topic...

Give me a bit more of a 'run-down' on your system (CPU, CPU cooler, RAM p/n, MOBO model and BIOS rev) and I can likely give you a 'workable' set of parameters to keep 'er 'humming'! Also, right above the above cited section, the Memtest Useage: section gives various RAM testing techniques...
 


I deleted easytune6 from pc and installed somethimg called Speccy instead. I would really appreciate a workable set of parameters if you have the time. My system consists of..... Core i7-930 cpu w/ stock cooler, Corsair Dominator 6GB DDR3-1600 (PC-12800) CL8 Memory p/n CM3X2G1600C8D, Gigabyte GA-X58A-UD5 Mobo Bios ver. F5, 2x nvidia quadro nvs 420's. Im going to try and test the ram later on tonight. Speccy is giving me all of the same values unlike Easytune, but im still having issues with performance. It only happens when i am receiving data during very fast markets(which just happens to be the worst possible time). My CPU never exceeds 21%, my memory use is usually between 36% and 41% and my network utilization is .01% to .03% whenever it happens. ?Im stumped? Could i have a disk read/write problem(i due cache a lot of data) or even GPU problem(i use 8 monitors)? Are there any other utilities out there i could use to troubleshoot??

One other thing.. i want to replace stock cpu fan, any recommendations?
 
I'll give a shot at numbers for you today yet - hopefully! For cooling, I usually recommend these; though the well-kept secret (obviously, they want people to spend as much as possible!) is that any aftermarket cooler will remove two to three times the heat that the pathetic stock Intel rotary postage stamp' can handle! You certainly don't need to get the 'latest, best, and, of course, most expen$ive'! For paste, I think these guys have the best on the market: TIM Consultants; TIM comparison testing; it's 'thixotropic' - becomes less viscous and 'flows' under shear forces, so when you 'wiggle down' the HSF, you get a great 'spread'; and, one of the things that impresses me - I think the last tube I bought cost me six bucks, and about seventy five cents shipping & handling - no 'plastic container in a blister pack in a fancy box in a pile of foam "ghost farts" in a larger UPS box! They throw it in a damn envelope, & mail it to you! [:bilbat:5] I fully expected the 'shipping & handling' to be more that the product! Always reminds me of an Andy Rooney I saw once: "Who's 'handling' this, anyway, and how much?" "Did they wash their hands first?" What if I want one that hasn't been 'handled'?"

While I've got your attention, gotta ask a question... Been thinking about a product - kind of a 'mini-supercomputer', specifically for futures traders, to do 'real-time' Black-Scholes analysis... What do you think people would pay? Does, say, ten to twelve thousand seem reasonable - that would be with vidcards (and, the price 'range' covers the differences there - would be four vidcards, that'd handle either six or twelve monitors, all together), but less the monitor array...
 
Well - sorry about 'no numbers' yet... Night before last, we had 4½ inches of rain, went into the basement, thanked god no water on floor around drainage sump. Yesterday, turned right, instead of left into the main basement, into a storage area - two ^%$# inches of water on the floor! Move everything, prop over to dry, vacuum up water, break back emptying ten gallon vac - on and on - by the time I got back up the stairs - directly to bed with a book (marcus Aurelius) as felt like had been beaten for a gang initiation!

I'm working on OC/numbers - what I would do at this point (will take some serious time, but you appear to already be losing time - and apparently, at the worst possible times!) is outlined here: Memory - Part IV - "Tweaking and tuning" You want to look about in the middle, where procedure to download and prep a copy of MemTest is given, and then toward the end, where actual testing techniques are listed - the one you want is: Pre-assembly DIMM 'qualification': for each stick, followed by: Full memory setup verification:

By the time you're done, I should have #'s for you - planning to do it three ways: stock w/1600 RAM, mild OC for use with current cooling, better OC for use once cooling is replaced...
 
...found another 'interesting' post: http://hardforum.com/showthread.php?t=1375251 - look familiar? Hmmm... Bothers me, too, that, although a google search for the RAM part number turns up plenty of reviews, a search at Corsair for the specs turns up - no such part number! Were my stuff - I'd contact 'em ( Corsair HelpDeask ) Again, my main concern, from the beginning, is that the SPD/XMP might be 'mis-programmed' in one or more of 'em - and I have absolutely no idea what the BIOS' setup routines will make of that!


GA-X58A-UD5
Intel i7-930 x21mult 2.8GHz .8-1.375V 130W TDP (to 3.36/4.0GHz)
Corsair Dominator 6GB DDR3-1600MHz (CM3X2G1600C8D): 8-8-8-24-?t 1.65v nominal

'Gentle' overclock (3.36 w/stock cooling) are in italics; faster OC (with any other cooling!) are in bold!

Uninstall any GIGABYTE 'system software': EasyTune 6, Dynamic Energy Saver, Smart 6, Auto Green...

If you haven't yet done it, start with a BIOS' "Load Optimized Defaults" with only three sticks of RAM in the white slots...

Before we start ramping things up, I want to teach you a new skill involving the BIOS: Do the <DEL> at the boot to enter the BIOS;
notice, at the bottom, the <F11> "Save CMOS to BIOS" - hit this, and you should get a menu that will show a number (the count varies by BIOS) of empty 'slots', each of which will store an entire set of BIOS parameters, to be re-loaded from the corresponding <F12> "Load CMOS from BIOS"; this is a wonderful overclocker's feature. What I do with it, is to save my 'baseline' working parameters, so if I change something that 'irritates' the board, and forces a reset of all the parameters to defaults, or, even worse, get so screwed up I need to do a 'clear CMOS', I can get back to my starting point with no effort, and without having to remember 85 separate settings! Another thing it prevents is two hours' troubleshooting, having forgotten a change to a crucial parameter - like, "wait a minute - didn't I have the Trd at seven?!" It's pretty self-explanatory, and I alway urge people to start right away by taking the time to give the 'slots' names that mean something: in two hours, "Try2" and "Try3" will not be very helpful, but "450@+10MCH" and "450@+15MCH" will! Another use is for 'green' settings; overclocks, as a rule, do not 'play well' with green features, such as 'down-clocking' and 'down-volting'; with the storage slots, you can set up one profile, say "Green", with all the settings at 'stock' values, and all the 'green' features enabled; another, say "Balls2Wall" with a full overclock, and all the 'green' stuff turned off... Another neat feature of this 'slot' system is, for most BIOS, the mechanism itself will keep track of which ones have booted successfully, and how many times (up to, I believe, a max of five)!

Now, the next step is the most time consuming - we've got to 'sidestep' XMP's limitations. The only reliable way to get the XMP timings for your RAM is to enable XMP - BUT - enabling XMP 'locks out' most all the other OCing parameter changes. Easy to get around, though. Having done a LoadOpt to set up the main config, the next thing you need to do is enable XMP, <F10> to save, exit, and reboot - and then run a pass of MemTest86+ (instructions for D/Ling and 'prepping' MemTest about the middle of Part IV - "Tweaking and tuning") to verify the the XMP for your DIMMs actually works! Assuming it's successful, go into the BIOS, and on the "Advanced Memory Settings" sub-page of the "MB Intelligent Tweaker(M.I.T.)" page of your BIOS, write down (preferable in a neat, tidy, easy-to-read table - extra points here for neatness!! Actually, neatness doesn't count toward your overclock, but accuracy does - if you make a mistake anywhere, you'll, guaranteed, have the devil's [:lorbat:8] own time finding it!) the "Channel Interleaving" and "Rank Interleaving" settings; then, enter the "Channel A Timing Settings" 'sub-sub-page' (?!), and write down all the settings. If I am correct, it will display two sets of timings - on the left, it will show you the 'LoadOpt's 'Auto' parameters; on the right, labeled either XMP or 'Profile (number?) settings', it will show you the 'fast RAM' parameters - these are the ones you need... As an example, here is a table from a client's machine (different platform...):

CAS 7
tRCD 8
tRP 7
tRAS 24

tRC 40
tRRD 5
tWTR 6
tWR 12
tWTP 23
tWL 7
tRFC 88
tRTP 6
tFAW 50
tCMD 2
static tRD 15



Next, 'back out', and go to the " Channel A Turnaround Settings" 'sub-sub-page' (?if they get many more 'layers' of menuing, I'm gonna lose it!), and scribble those down as well. Now back out again, and look at both sets of 'channel B's' & 'channel C's' - they should be the same - just double check, and note anything differing...

On to the 'quick' part! First, disable XMP. Then, while you're still 'familiar' with those memory pages, go back and set all of 'em - to the values you wrote down - and don't forget to check the two on the 'main page', "Channel Interleaving" and "Rank Interleaving"... To set these, start with:

"Performance Enhance" to "Standard"
"DRAM Timing Selectable (SPD)" to "Expert"...

Once they're all set -

...next, two adjustments for the fact that you are running two DIMMs per channel - XMP values are 'figured' for a single module per channel:
Set "Command Rate(CMD)" to "2", "2T", or "2N" (whichever way your BIOS shows it...)
Take the number you wrote down for the tRFC, multiply it by 1.15, round up to the next highest integer, and set tRFC to the resultant value...


At this point, I usually do an <F11> save to BIOS, just so I can easily 'get back to where I was', calling it something like "RAMisSet"...


On the "MB Intelligent Tweaker(M.I.T.)" page:

"CPU Clock Ratio" should already be at "X21"
"CPU Frequency" - this one can't be set, it's calculated

On the "Advanced CPU Features" sub-page (there we go again![:fixitbil:9] ):

"Intel(R) Turbo Boost Tech." to "Enabled"
"CPU Enhanced Halt (C1E)" to "Disabled"
"C3/C6/C7 State Support" to "Disabled"
"CPU Thermal Monitor" to "Enabled
"CPU EIST Function" to "Disabled"
"Bi-Directional PROCHOT" to "Enabled"
"Virtualization Technology" to "Enabled", if you intend to use it - it has costs...


On the "UnCore & QPI Features" sub-page:

"QPI Clock Ratio" to "X48"
"QPI Clock Ratio" to "X44"
"QPI Clock Ratio" to "X36"
"Uncore Clock Ratio" to "X25"
"Uncore Clock Ratio" to "X21"
"Uncore Clock Ratio" to "X17"
"Isochronous Support" to "Enabled"

...under "Standard Clock Control":

"Base Clock(BCLK) Control" to "Enabled"
"BCLK Frequency(Mhz)" to "133"
"BCLK Frequency(Mhz)" to "160"
"BCLK Frequency(Mhz)" to "190"

...under "Advanced Clock Control":

...should all be OK as set...


On the "Advanced DRAM Features" sub-page"

"System Memory Multiplier" to "X12"
"System Memory Multiplier" to "X10"
"System Memory Multiplier" to "X8"


On the "Advanced Voltage Settings" sub-page:

"Load-Line Calibration" is anyone's guess - this seems to work differently on every single board I do - even, sometimes, differing between boards of the same model with different CPUs! My best guess - start with it at "Standard", and, maybe, adjust when 'running thermals'...
"CPU Vcore" to "1.325V" - idea here is to start high, to get 'er stable first, then reduce voltage when 'running thermals' - doing it the other way - starting low, and trying to 'creep up' the Vcore is a real PITA, as, often, if the MOBO's not 'happy' with OC settings, she'll 'crank back' to the LoadOpt parameters - and you lose anything you didn't have the foresight to 'save to BIOS'!

"CPU Vcore" to "Auto"
"CPU Vcore" to "1.2375V"
"CPU Vcore" to "1.300V"
"QPI/Vtt Voltage" to "1.250V"
"PCH Core" to "1.10V"
"DRAM Voltage" to "1.68V" - 'even' increments only - need the slightly higher Vdimm as you are running two DIMMs per channel...


On the "Advanced BIOS Features" page:

"Quick Boot" to "Disabled"
"Full Screen LOGO Show" to "Disabled" (easier to get into the BIOS + lets you 'see' the POST...

On the "Integrated Peripherals" page:

...disable everything you're not actually planning on using - no 1394 port - turn it off!; same for serial port, eSATA, etc - if you don't need it, the processor/system doesn't need the 'overhead' either!

On the "Power Management Setup" page:

"ACPI Suspend Type" to "S1(POS)" (S3 is notoriously difficult to get working properly on OC'd boards...)

On the "PC Health Status" page:

"CPU Warning Temperature" to "70C"

And that should do it! Time for, first, an <F11> 'save to BIOS', followed by an <F10> save & exit - power down, add the other three sticks of RAM - hopefully, she'll reboot, and WORK!

I should point out that getting two reboots in a row here is perfectly normal behavior; it seems that, when you change certain settings (and we don't exactly know which ones - the only sure one I know is Trd - if you change it, I think you get the 'twin' reboot) it boots once to 'see where it's at', recalculates its remaining 'auto' settings, saves them, and then boots again. Three reboots in a row, however, usually indicates that the board was 'given indigestion' by your settings, and is going back to defaults. This sometimes goes astray, and it doesn't get back into proper operation - for example, at this point, mine will sometimes 'lock' itself into 111MHz x a six multiplier - and take a week to do a whole boot - that's time to do a CMOS reset, and use your 'stored' <F12> profile to get back to where you were...

Good luck!

Bill
 



Ill ask around about the 'real time' Black and Scholes machine. i know a few option traders from my NYMEX days. Options are still the one thing from my exchange that are still an active open outcry market. The transition to electronic trading for options has yet to fully take place. I believe its because of the complications of calculating everything through the servers. So most guys have there own computers on the trading floor doing all of the calculations.
 
i didnt have a chance to come into the office over the weekend to run memtest at all. Ill set one dimm up tonight and run it. I actually found the packaging for my memory last night and im realizing now that the box shows 'TR3X6G1600C8D G' not 'CM3X2G1600C8D' like easytune and Specy are showing. Im not sure if its because it was a memory kit and if thats why the numbers for mem and for the actually kit are different. I bought everything at microcenter and it took there 1yr replacement plan. I wonder if i should just go and exchange it for another kit of equal value and not even worry about running the tests. Although i am curious to see if there is a problem with one or all of the mem sticks.

One other thing, about that post u found. I basically see the same part numbers he was seeing and i also installed the same set he mentions (its that number i found on the memory's packaging). So im guessing that the kit number and the actual part numbers of the ram are different. It definitely looks im having the same problem he had though.
 
Since ive been using Speccy and got rid of EasyTune6 im seeing these settings for ram without me doing anything to system. Does anything look out-of-wack to u.

Memory
Type DDR3
Size 6144 MBytes
Channels # Triple
DRAM frequency 539.7 MHz
CAS# Latency (CL) 8 clocks
RAS# to CAS# delay (tRCD) 8 clocks
RAS# precharge (tRP) 8 clocks
Cycle time (tRAS) 20 clocks
Command rate (CR) 1T
SPD
Number of SPD modules 3
Slot #1
Type DDR3
Size 2048 MBytes
Manufacturer Corsair
Max bandwidth PC3-10700H (667 MHz)
Part number CM3X2G1600C8D
Week/year 13 / 09
SPD Ext. EPP
XMP-6925200
Frequency 3462600 MHz
JEDEC #3
Frequency 666.7 MHz
CAS# latency 9.0
RAS# to CAS# 10
RAS# Precharge 10
tRAS 25
tRC 34
Voltage 1.500 V
JEDEC #2
Frequency 592.6 MHz
CAS# latency 8.0
RAS# to CAS# 8
RAS# Precharge 8
tRAS 22
tRC 30
Voltage 1.500 V
JEDEC #1
Frequency 444.4 MHz
CAS# latency 6.0
RAS# to CAS# 7
RAS# Precharge 7
tRAS 17
tRC 23
Voltage 1.500 V

Im seeing the same for all 3 slots. the only thing different for all 3 is whats in bold above
 
Thanx for putting the time in and coming up with those settings for me. I really appreciate it. After testing ram ill give them a try. From what u said earlier it looks like it going to take me a few days to test everything. Unless u think i should use my replacement plan and scratch the tests?

Your post....Memory - Part IV - "Tweaking and tuning" is great . It explains so much. Thanx again for everything
 
Are these hard drive counts that Speccy is showing me acceptable. Im not sure what im looking at.....
Hard Drives
WDC WD1002FAEX-00Z3A0 ATA Device
Manufacturer Western Digital
Form factor TB/3.5-inch (new format)
Business Unit/Brand Desktop/WD Caviar®
RPM/Buffer Size or Attribute 10,000 RPM with 16 MB cache
Interface IDE
Capacity 1000GB
Real size 1,000,200,658,432 bytes
S.M.A.R.T
01 Read Error Rate 200 (200 worst) Data 0000000000
03 Spin-Up Time 172 (170) Data 000000111F
04 Start/Stop Count 100 (100) Data 000000008F
05 Reallocated Sectors Count 200 (200) Data 0000000000
07 Seek Error Rate 100 (253) Data 0000000000
09 Power-On Hours (POH) 099 (099) Data 000000058A
0A Spin Retry Count 100 (100) Data 0000000000
0B Recalibration Retries 100 (100) Data 0000000000
0C Device Power Cycle Count 100 (100) Data 0000000070
C0 Power-off Retract Count 200 (200) Data 0000000030
C1 Load/Unload Cycle Count 200 (200) Data 000000005E
C2 Temperature 116 (107) Data 000000001F
C4 Reallocation Event Count 200 (200) Data 0000000000
C5 Current Pending Sector Count 200 (200) Data 0000000000
C6 Uncorrectable Sector Count 200 (200) Data 0000000000
C7 UltraDMA CRC Error Count 200 (200) Data 0000000000
C8 Write Error Rate / Multi-Zone Error Rate 200 (200) Data 0000000000
Temperature 96 °F
Temperature range ok (less than 122 °F)
Status Good
Partition 0
WDC WD3000GLFS-01F8U0 ATA Device
Manufacturer Western Digital
Form factor GB/2.5- to 3.5-inch adapter
Business Unit/Brand Enterprise/WD VelociRaptor®
Interface/Connector SATA 3 Gb/s with 22-pin SATA connector/SATA 1.5 Gb/s with 22-pin SATA connector (Mobile)
Interface IDE
Capacity 300GB
Real size 300,069,052,416 bytes
S.M.A.R.T
01 Read Error Rate 149 (144 worst) Data 0000006AE7
03 Spin-Up Time 196 (196) Data 0000000C5E
04 Start/Stop Count 100 (100) Data 000000006C
05 Reallocated Sectors Count 200 (200) Data 0000000000
07 Seek Error Rate 200 (200) Data 0000000000
09 Power-On Hours (POH) 099 (099) Data 0000000591
0A Spin Retry Count 100 (100) Data 0000000000
0B Recalibration Retries 100 (100) Data 0000000000
0C Device Power Cycle Count 100 (100) Data 000000006B
C0 Power-off Retract Count 200 (200) Data 000000002A
C1 Load/Unload Cycle Count 200 (200) Data 000000006C
C2 Temperature 119 (112) Data 000000001C
C4 Reallocation Event Count 200 (200) Data 0000000000
C5 Current Pending Sector Count 200 (200) Data 0000000000
C6 Uncorrectable Sector Count 200 (200) Data 0000000000
C8 Write Error Rate / Multi-Zone Error Rate 200 (200) Data 0000000000
Temperature 88 °F
Temperature range ok (less than 122 °F)
Status Good
Partition 0
Partition 1

I dont even have HDD S.M.A.R.T. Capability turned on in bios, i guess Speccy can calculate all by itself.


Sorry for bombarding you with all these posts today
 
There's only a few that really matter - at least IMHO; Reallocated Sectors Count, Seek Error Rate, Reallocation Event Count, and, likely, Uncorrectable Sector Count... The seek errors indicate your drive is starting to have difficulties 'finding where it's at', locating the read/write heads over the data tracks accurately; the remainder pretty much say the drive's developing 'unuseable' spots - if these start climbing, there's a reason - drive failing, power problems, what-have-you... At any rate, they're still zeroed on your drives, so you're in good shape!

Either way, I recommend 'qualification' testing of RAM, before putting it 'into service'...
 
i did a load optimized yest then turned on xmp and ran memtest last night. when it booted from cd i just let it run. i didnt change any settings. Did i do it correctly? Then today when i got to work memtest said it passed, but it was still running. Im doing 'qualification' testing of each individual stick now. How many passes do u recommend?
 
I usually let each stick go 'a shift': I'll start one toward evening, let it run overnight; then, next morning, check results and start another, let it run till early afternoon; again, check results, and let the next test run 'till evening. As most of my clients are building [strike]cheap[/strike] less expensive machines these days, I'm usually working with 'two stick' installs, so that means I can get in individual testing for each stick, and 'full setup' qualification in roughly one twenty-four hour day... When I did the RAM for my workstation (four DIMMs), I bought a few extra sticks to select the four fastest, did quite a bit of 'push' testing (seeing if, and how high, they'd overclock), and spent about three weeks on DIMM testing and qualification alone!
 
A few questions
1. I let one run all night last night came in today and it had done 44 passes with no errors. Is just letting Memtest run the correct approach or should i be doing anything in the configuration screen?

2. Whenever i do a load optimized bios keeps shutting off 64 bit and truning on 32 bit. Should i change this setting before i go and write down all of my timings or is it ok to do this at the end.

3. About the settings u gave me.... I dont have a PCH core setting I only have IOH core and ICH core should i set both of those to "1.10v" instead?

4. and for QPI/Vtt voltage i dont have "1.250V" mine has "1.255V" is that ok?

5. Also, Round trip latency for Channel A , B and C were the only things with different values. A was 55 and B was 58 and C was 60. Should they be the same? Or is that because it takes a little bit longer to reach each Dimm slot because each one is a little bit further away.

One other thing when i go into Channel A timing settings sub menu. I only see settings on left and to the right all 'AUTOS'. But when i shut enable/disable XMP all of the setting on the left do change. I never see both sets at the same time like you thought.
 
1 - I never configure anything! Works fine 'hands off'...

2 - If you're talkiking about the "HPET mode" on the "Power Management Setup" page - doesn't matter until you actually install OS - and, really, doesn't matter all that much then! I can't think of much else denoted as '32' vs '64' bit...

3 - My sloppiness! 1156's use a PCH (platform controller hub); 1366's use an IOH/ICH pair - set the IOH to the 1.1V, leave the ICH where it's at...

4 - Anyplace between, likely, 1.225 and 1.255 - as low as it'll go and still work (and, best way is to try it high, and lower it - then you know why it quit working, and can 'back up' one step...

5 - set these as shown in your BIOS - they are established by the memory controllers 'training' process at startup, and it's damned difficult to improve on 'em...